Snowflake Decision hits 57.63, top of 72 models

Snowflake's decision model is meant for bounded, repeated questions — which category a record belongs to, how risky a transaction is, whether a ticket needs more work — rather than open-ended reasoning. The company has positioned it as another option in Cortex AI Functions, next to the general-purpose models already offered, and says it's suited to running across thousands of rows where speed and cost matter. In the vendor's own comparison, Snowflake Decision scored 57.63, ahead of the other 71 models on the same 29 benchmarks. Snowflake notes that current interface limits, such as a maximum of 32 options, kept it from running the full Hugging Face benchmark, and that every model's score was calculated from the same benchmark results using the index's published, chance-adjusted methodology.